String delimited regular expression

Question

I'm trying to build a bbcode parser, but I'm having quite some problems figuring out how to avoid matching too widely. For example I want to implement a [list] to conversion like this:

\[list\](.*)\[/list\]

would be replaced by this:

<ul>$1</ul>

This works fine, except if I have two lists where the regular expression matches the beginning tag of the first list and the ending tag of the second. So this

[list]list1[/list] [list]list2[/list]

becomes this:

<ul>list1[/list] [list]list2</ul>

which produces really ugly output. Any idea on how to fix this?

Answer 1

The method you're using may not end up being a particularly good approach, but to solve that specific problem, just change to non-greedy matching:

\[list\](.*?)\[\/list\]

Note that this way will have trouble with nested lists instead of back-to-back ones.

Answer 2

If what you are doing is not just a lightweight hack, but something more permanent, you probably want to move to a real parser. Regexps in Java are particularly slow (even with precompiled patterns) and matching nested constructs (especially different nested contructs like "foo [u][i] bar [s]baz[/s][/i][/u]" ) is going to be a royal pain.

Instead, try using a state-based parser, that repeatedly cuts your sentence in sections like "foo " / (u) / "[i] bar [s]baz[/s][/i][/u]", and maintains a set of states that flip whenever you encounter the matching construct delimiter.